70
by Michael Hope
Added the C only routines from GLIBC 2.16+20120607~git24a6dbe |
1 |
/* Copyright (C) 1991,1993-1997,1999,2000,2003,2006
|
2 |
Free Software Foundation, Inc.
|
|
3 |
This file is part of the GNU C Library.
|
|
4 |
Based on strlen implementation by Torbjorn Granlund (tege@sics.se),
|
|
5 |
with help from Dan Sahlin (dan@sics.se) and
|
|
6 |
bug fix and commentary by Jim Blandy (jimb@ai.mit.edu);
|
|
7 |
adaptation to strchr suggested by Dick Karpinski (dick@cca.ucsf.edu),
|
|
8 |
and implemented by Roland McGrath (roland@ai.mit.edu).
|
|
9 |
||
10 |
The GNU C Library is free software; you can redistribute it and/or
|
|
11 |
modify it under the terms of the GNU Lesser General Public
|
|
12 |
License as published by the Free Software Foundation; either
|
|
13 |
version 2.1 of the License, or (at your option) any later version.
|
|
14 |
||
15 |
The GNU C Library is distributed in the hope that it will be useful,
|
|
16 |
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
17 |
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
18 |
Lesser General Public License for more details.
|
|
19 |
||
20 |
You should have received a copy of the GNU Lesser General Public
|
|
21 |
License along with the GNU C Library; if not, see
|
|
22 |
<http://www.gnu.org/licenses/>. */
|
|
23 |
||
24 |
#include <string.h> |
|
25 |
#include "memcopy.h" |
|
26 |
#include <stdlib.h> |
|
27 |
||
28 |
#undef strchr
|
|
29 |
||
30 |
/* Find the first occurrence of C in S. */
|
|
31 |
char * |
|
32 |
strchr (s, c_in) |
|
33 |
const char *s; |
|
34 |
int c_in; |
|
35 |
{
|
|
36 |
const unsigned char *char_ptr; |
|
37 |
const unsigned long int *longword_ptr; |
|
38 |
unsigned long int longword, magic_bits, charmask; |
|
39 |
unsigned char c; |
|
40 |
||
41 |
c = (unsigned char) c_in; |
|
42 |
||
43 |
/* Handle the first few characters by reading one character at a time.
|
|
44 |
Do this until CHAR_PTR is aligned on a longword boundary. */
|
|
45 |
for (char_ptr = (const unsigned char *) s; |
|
46 |
((unsigned long int) char_ptr & (sizeof (longword) - 1)) != 0; |
|
47 |
++char_ptr) |
|
48 |
if (*char_ptr == c) |
|
49 |
return (void *) char_ptr; |
|
50 |
else if (*char_ptr == '\0') |
|
51 |
return NULL; |
|
52 |
||
53 |
/* All these elucidatory comments refer to 4-byte longwords,
|
|
54 |
but the theory applies equally well to 8-byte longwords. */
|
|
55 |
||
56 |
longword_ptr = (unsigned long int *) char_ptr; |
|
57 |
||
58 |
/* Bits 31, 24, 16, and 8 of this number are zero. Call these bits
|
|
59 |
the "holes." Note that there is a hole just to the left of
|
|
60 |
each byte, with an extra at the end:
|
|
61 |
||
62 |
bits: 01111110 11111110 11111110 11111111
|
|
63 |
bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD
|
|
64 |
||
65 |
The 1-bits make sure that carries propagate to the next 0-bit.
|
|
66 |
The 0-bits provide holes for carries to fall into. */
|
|
67 |
switch (sizeof (longword)) |
|
68 |
{
|
|
69 |
case 4: magic_bits = 0x7efefeffL; break; |
|
70 |
case 8: magic_bits = ((0x7efefefeL << 16) << 16) | 0xfefefeffL; break; |
|
71 |
default: |
|
72 |
abort (); |
|
73 |
}
|
|
74 |
||
75 |
/* Set up a longword, each of whose bytes is C. */
|
|
76 |
charmask = c | (c << 8); |
|
77 |
charmask |= charmask << 16; |
|
78 |
if (sizeof (longword) > 4) |
|
79 |
/* Do the shift in two steps to avoid a warning if long has 32 bits. */
|
|
80 |
charmask |= (charmask << 16) << 16; |
|
81 |
if (sizeof (longword) > 8) |
|
82 |
abort (); |
|
83 |
||
84 |
/* Instead of the traditional loop which tests each character,
|
|
85 |
we will test a longword at a time. The tricky part is testing
|
|
86 |
if *any of the four* bytes in the longword in question are zero. */
|
|
87 |
for (;;) |
|
88 |
{
|
|
89 |
/* We tentatively exit the loop if adding MAGIC_BITS to
|
|
90 |
LONGWORD fails to change any of the hole bits of LONGWORD.
|
|
91 |
||
92 |
1) Is this safe? Will it catch all the zero bytes?
|
|
93 |
Suppose there is a byte with all zeros. Any carry bits
|
|
94 |
propagating from its left will fall into the hole at its
|
|
95 |
least significant bit and stop. Since there will be no
|
|
96 |
carry from its most significant bit, the LSB of the
|
|
97 |
byte to the left will be unchanged, and the zero will be
|
|
98 |
detected.
|
|
99 |
||
100 |
2) Is this worthwhile? Will it ignore everything except
|
|
101 |
zero bytes? Suppose every byte of LONGWORD has a bit set
|
|
102 |
somewhere. There will be a carry into bit 8. If bit 8
|
|
103 |
is set, this will carry into bit 16. If bit 8 is clear,
|
|
104 |
one of bits 9-15 must be set, so there will be a carry
|
|
105 |
into bit 16. Similarly, there will be a carry into bit
|
|
106 |
24. If one of bits 24-30 is set, there will be a carry
|
|
107 |
into bit 31, so all of the hole bits will be changed.
|
|
108 |
||
109 |
The one misfire occurs when bits 24-30 are clear and bit
|
|
110 |
31 is set; in this case, the hole at bit 31 is not
|
|
111 |
changed. If we had access to the processor carry flag,
|
|
112 |
we could close this loophole by putting the fourth hole
|
|
113 |
at bit 32!
|
|
114 |
||
115 |
So it ignores everything except 128's, when they're aligned
|
|
116 |
properly.
|
|
117 |
||
118 |
3) But wait! Aren't we looking for C as well as zero?
|
|
119 |
Good point. So what we do is XOR LONGWORD with a longword,
|
|
120 |
each of whose bytes is C. This turns each byte that is C
|
|
121 |
into a zero. */
|
|
122 |
||
123 |
longword = *longword_ptr++; |
|
124 |
||
125 |
/* Add MAGIC_BITS to LONGWORD. */
|
|
126 |
if ((((longword + magic_bits) |
|
127 |
||
128 |
/* Set those bits that were unchanged by the addition. */
|
|
129 |
^ ~longword) |
|
130 |
||
131 |
/* Look at only the hole bits. If any of the hole bits
|
|
132 |
are unchanged, most likely one of the bytes was a
|
|
133 |
zero. */
|
|
134 |
& ~magic_bits) != 0 || |
|
135 |
||
136 |
/* That caught zeroes. Now test for C. */
|
|
137 |
((((longword ^ charmask) + magic_bits) ^ ~(longword ^ charmask)) |
|
138 |
& ~magic_bits) != 0) |
|
139 |
{
|
|
140 |
/* Which of the bytes was C or zero?
|
|
141 |
If none of them were, it was a misfire; continue the search. */
|
|
142 |
||
143 |
const unsigned char *cp = (const unsigned char *) (longword_ptr - 1); |
|
144 |
||
145 |
if (*cp == c) |
|
146 |
return (char *) cp; |
|
147 |
else if (*cp == '\0') |
|
148 |
return NULL; |
|
149 |
if (*++cp == c) |
|
150 |
return (char *) cp; |
|
151 |
else if (*cp == '\0') |
|
152 |
return NULL; |
|
153 |
if (*++cp == c) |
|
154 |
return (char *) cp; |
|
155 |
else if (*cp == '\0') |
|
156 |
return NULL; |
|
157 |
if (*++cp == c) |
|
158 |
return (char *) cp; |
|
159 |
else if (*cp == '\0') |
|
160 |
return NULL; |
|
161 |
if (sizeof (longword) > 4) |
|
162 |
{
|
|
163 |
if (*++cp == c) |
|
164 |
return (char *) cp; |
|
165 |
else if (*cp == '\0') |
|
166 |
return NULL; |
|
167 |
if (*++cp == c) |
|
168 |
return (char *) cp; |
|
169 |
else if (*cp == '\0') |
|
170 |
return NULL; |
|
171 |
if (*++cp == c) |
|
172 |
return (char *) cp; |
|
173 |
else if (*cp == '\0') |
|
174 |
return NULL; |
|
175 |
if (*++cp == c) |
|
176 |
return (char *) cp; |
|
177 |
else if (*cp == '\0') |
|
178 |
return NULL; |
|
179 |
}
|
|
180 |
}
|
|
181 |
}
|
|
182 |
||
183 |
return NULL; |
|
184 |
}
|